Authoritative working tracker of source materials not yet synthesized into a Published / Draft (or Publications+) card. Snapshot 2026-06-09. Line-level lists live in the repo archive files (below); this card is the working dashboard the team works from.
Run record + future work: the 2026-06-21 canonical-sheet OA ingestion pass (82 new RawData cards) and the blocked tail / decisions-needed are documented in the sibling card +Canonical Sheet OA Ingestion 2026-06-21 and Future Work (9256).
Archive snapshots (local repo, dated):
- scripts/archive/publications_coverage_audit_2026_06_09.txt — coverage audit (canonical sheet vs wiki vs publication_texts/).
- scripts/archive/sources_pending_analysis_2026_06_09.txt — full line-level backlog feeding this tracker.
Baseline counts: canonical sheet 348 rows · wiki 214 RawData+Publications + 56 Publications paper cards · publication_texts/: json 24 / ingest 53 / pdfs 53 / mattermost_papers 154 / readai 10.
| # | Bucket | Count | Status |
|---|---|---|---|
| 1 | Canonical-sheet pubs missing a card | 144 (55 curated 2019-2025 + 89 scholar 1990-2000) | pending |
| 2 |
mattermost_papers/ un-synthesized |
~145 of 154 | pending |
| 3 | readai_reports/ |
9 | deferred (process artifacts, not pubs) |
| 4 |
ingest/ + pdfs/ un-synthesized |
~28 (the ~30 arxiv_*.pdf are ALL already carded — see P1) |
partial — raw layer mostly done |
| 5 |
RawData+Publications not yet synthesized to a Publications card |
~155 of 214 | pending (overlaps #2) |
arxiv_*.pdf — ✅ DONE 2026-06-09: all 30 already have RawData+Publications cards (4 also have Publications synthesis cards); 0 missing. The “ID-named PDF” flag was a filename↔title matching artifact, not a real gap. 26 of 30 still lack a Publications synthesis card → folded into P2 / Bucket 5. Reconcile table: scripts/archive/arxiv_reconcile_2026_06_09.txt.Publications+MORK Research Family (Draft) mapping each paper/version’s contribution (AdaptiMORK vs QuantiMORK vs Miner vs Tensor_Networks vs slots vs theory…), cross-linked to the RawData source cards + existing MORK algorithm/knowledge cards (kept separate from those functional cards). Review loop = lighter-touch: Claude drafts from RawData + prior findings → Codex single-turn review for factual drift / missing caveats → publish; Gemini optional high-level framing only (NOT the full brief→extraction→reconciliation cluster-pilot machinery). MORK becomes the template for the other families. STATUS 2026-06-09: pilot card PUBLISHED (Draft) — Publication Maps+MORK Research Family (8441 + tag 8442); Codex factual-drift pass cleared (5 fixes applied), Gemini synthesis sign-off. ~19 MORK working-papers mapped across 5 sub-families. Template now set for the remaining families (MeTTa specs, Weakness/singular-learning, FabricPC/predictive-coding, Causal Coding, concept-blending/quantale, SubRep, Omegaclaw PLN, PyMeTTa).
Year/venue schema implemented + arxiv backfill COMPLETE (2026-06-21, bulk-ingestion pipeline agent) — added real subcards +year (Number) / +venue (Phrase) / +metadata_source (Phrase: curated_sheet |
arxiv | inferred) to RawData+Publications cards; backfilled all 30 of 30 arxiv cards (year + venue=arXiv + metadata_source=arxiv; 90 subcards, 0 errors). The 2 final cards (3226 ActPC-Geom, 3235 Distinction Graphs) were manually verified by arxiv-id-in-content (https://arxiv.org/abs/2501.04832,scripts/archive/arxiv_backfill_manifest_2026_06_21.txt /abs/1902.00741) — not fuzzy score — then backfilled. Manifest: . scripts/ingest_publications.rb updated to idempotently emit the 3 subcards on future ingests: provenance-aware (rank curated_sheet>arxiv>inferred; inferred never overwrites a curated value), logs every proposed write, METADATA_DRY_RUN=1 for log-only; arxiv records (URL or explicit field) emit venue=arXiv/metadata_source=arxiv. Dry-run validated against live deck (UNCHANGED/SKIP/PROPOSE-UPDATE paths). publication_texts/arxiv_metadata.json stamped venue=arXiv/metadata_source=arxiv. Intentional recorded exception: the +year/+venue/+metadata_source subcards are PUBLIC (*all+*read=Anyone) — non-sensitive bibliographic metadata, gated RawData full-text parent unaffected. Addresses the PRIMARY (year/venue) handoff ask end-to-end. (Matcher note: cards.name does NOT store the +joined compound name on this deck — match via Card.fetch + cardname.parts.) |
Publication Maps+MORK Research Family (8441, Draft + tag 8442, ai_generated). Family map over ~19 MORK Mattermost working-papers (Jul–Oct 2025), 5 sub-families (Core theory/indexing · PLN execution · Algorithm substrate · WILLIAM compression · Neural/quantitative), version chains collapsed (PLN FactorGraph v3.2 / AdaptiMORK v8 / WILLIAM formal paper), implementation-reality-check section grounding against the AtomSpace-pilot codebase findings. Cross-linked to RawData sources + functional MORK/MOSES/PLN/QuantiMork cards (kept separate). Codex pass cleared; Gemini signed off. This is the template for the other Mattermost research families.
RawData+Publications+Autonomous Intelligent Reinforcement Inferred Symbolism (8150) + +tag (8151). Closed the one cluster-pilot-discovered pub that lacked a card.arxiv_*.pdf are already carded; the apparent “carded-by-ID-only” gap was a matching artifact. No cards created. scripts/archive/arxiv_reconcile_2026_06_09.txt.The canonical Google Sheet (1Wf1X..., “SNET - Research Publication Catagorization”) was located 2026-06-21 and re-pulled via XLSX. It now spans 1990–2025 continuously (786-row “Research Publications, Full List” tab; the old “2001–2018 absent” gap is CLOSED) plus a 139-row “2017 to 2024” tab = 925 bibliographic rows. Read-only reconciliation manifest built (Codex-scoped): 925 rows → 746 distinct (179 dups) vs 215 existing RawData+Publications cards (Decko-aware match). Classification: 42 EXISTS / 524 MISSING_CAN_FETCH_TEXT / 175 MISSING_NO_TEXT_SOURCE / 5 AMBIGUOUS / 179 DUPLICATE. Artifacts (local, gitignored + externally backed up): scripts/archive/snet_pubs_reconcile_manifest_2026_06_21.txt (+ classified TSV, review resolutions, backfill plan, backfill audit).
Existing-card metadata backfill DONE 2026-06-21 (refined policy, Codex-approved): 10 cards written — 8 non-arxiv missing-year backfills (year + metadata_source=curated_sheet; 3155/3161/3172/3218/3512/3524/3710/8150) + 2 venue upgrades to the real AGI-2020 venue (3231/3307, arxiv year kept). 28 deliberately skipped with reasons: 22 arxiv cards left at metadata_source=arxiv (year already matches — no churn, arxiv is more precise); 1 year-conflict flagged for review (3269: arxiv 2020 vs sheet 2019); 3 DO_NOT_BACKFILL (3246 sheet-year=arxiv-id error / 3489 conflicting years / 3558 informal design doc); truncated venue fragments not written. 0 errors, all verified. Audit: scripts/archive/snet_pubs_backfill_audit_2026_06_21.txt.
Step 3 fetch/verify probe + Batch 1 creates DONE 2026-06-21: read-only fetch probe calibrated retrievability (“MISSING_CAN_FETCH_TEXT” = URL present, NOT necessarily retrievable): only arXiv reliably returns full text; publisher DOIs (Springer/Wiley/Tandfonline/JSTOR/ScienceDirect) are paywalled/403; ResearchGate 403; some OA repos 404. Host breakdown of the 524: 40 arxiv(open) / 31 repo-OA / 231 publisher-DOI(paywall) / 41 publisher-direct / 172 other. Batch 1 (arXiv) COMPLETE: 35 RawData+Publications cards CREATED with full PDF body + +year/+venue=arXiv/+metadata_source=arxiv (35/35 verified; canonical naming; chunked at 50k; 1 needed null-byte sanitization). 4 AMBIGUOUS (arXiv-id-in-body = citations) left for manual confirm; 1 EXISTS_NOOP. Audit batch1_create_audit_2026_06_21.txt. Batch 2 NEXT (read-only): Unpaywall/OpenAlex/Semantic Scholar/goertzel.orgOA-resolver pass over the 231 publisher-DOI rows, classified by ACTUAL retrievable full text. NO abstract-only cards (separate policy decision). Paid access: institutional/TDM APIs only, no personal-subscription scraping.
Batch 2 OA-resolver pass DONE 2026-06-21 (read-only): 230 publisher-DOI rows through Unpaywall→OpenAlex→Semantic Scholar (DOI-first, cached, rate-limited). Result: 10 OA_FULLTEXT_READY / 5 abstract-only / 7 no-OA / 203 PAYWALL_BLOCKED / 5 fetch-failed. Free OA recovered only ~4% — the publisher back-catalog is overwhelmingly paywalled (institutional/TDM access is the only bulk route). Server dup gate → 8 OA full-text CREATE_READY (4 Frontiers cc-by, 3 goertzel.org,8(Codex 1 sciendo, 1 institutional repo) + 1 ambiguous (Lojban++ DOI in card 8793) + 1 excluded (proceedings front-matter). OA full-text cards CREATED + verified 2026-06-21 schema: +year from sheet, +venue canonical journal, +metadata_source=curated_sheet, body provenance in header+audit). 4 Frontiers + 1 sciendo JAGI + 1 ACM-SAC + 2 goertzel.org/IJMC.Cumulative Lojban++ held (ambiguous), front-matter excluded. 2026-06-21: 43 new RawData cards (35 arXiv Batch 1 + 8 OA Batch 2).
Confirm-then-create cleanup DONE 2026-06-21: 4 arXiv-ambiguous confirmed-distinct + created (citations in other cards); Lojban++ deferred on “+”-in-name mechanics (needs slug convention); 3 DISTINCT_MISSING kept out (no OA body). Batch 3 probe DONE 2026-06-21 (read-only): 244 non-arXiv/no-DOI rows → 42 OA_FULLTEXT_READY (all title-NEW) / 95 abstract-only / 44 paywalled / 41 fetch-failed / 20 ambiguous. The 42 are NOT clean auto-create: 22 title-unverified (need title-in-body check) + 6 informal-host docs (hold) + 1 lower-conf. Create-prep pass proposed before writing the clean ~30-35.
Batch 3 creates DONE 2026-06-21: create-prep verified the 42 (title-in-body), created 34 (full body, +year/+metadata_source=curated_sheet, +venue on 7 clean venues only; 6 informal-host docs excluded, 2 ambiguous held). Lojban++ created as “Lojban Plus Plus” (+display_title).
CUMULATIVE 2026-06-21: 82 new RawData+Publications cards (35 arXiv + 8 OA + 4 cleanup + 1 Lojban + 34 OA); total title-cards 215→297. Plus metadata backfills: 30 arXiv + 10 sheet.
STILL PENDING: (1) Lojban++ slug convention decision; (2) Batch 3 create-prep + create the clean ~30-35 (after sign-off); (3) the ~203 paywalled + ~175 NO_TEXT_SOURCE + 95 abstract-only + 44 Batch-3-paywalled remain BLOCKED pending institutional/TDM access or an explicit abstract-only policy change (both parked per Codex) (each needs retrievable body text from FullTextURL/ArticleURL/DOI — no metadata-only cards); 175 MISSING_NO_TEXT_SOURCE stay blocked; + 3 DISTINCT_MISSING reclassified create-candidates (PLN-temporal-procedural 2023, PLN 2014 EGI, AGI-preschool 2014).
Publications tree but years are inferred, not stored as card metadata.+year (Number) / +venue / +metadata_source on RawData+Publications; all 30 arxiv cards backfilled; ingest_publications.rb now emits the 3 fields idempotently on every ingest (provenance-aware, won’t overwrite curated). Still pending: broader 2001-2018 era backfill (~155 cards, inferred years) and the 54 curated creates — both BLOCKED on canonical-sheet location (venue/DOI) + source full-text.Generated by the MeTTa-runtime-pilot session’s publications coverage audit (Task #3), 2026-06-09.